Upgraded computer; Now all my games crash!

[antdude]

Guys, check out my CMOS settings (with L2 cache options enabled and unstable system):

SDRAM RAS-to-CAS delay = 3 (tried 2, but no effect)
SDRAM RAS PreCharge time = 3 (tried 2, but no effect)
SDRAM CAS Latency Time = 3 (tried 2, but no effect)
SDRAM PreCharge Control = Enabled
DRAM Data Integrity Mode = Non-ECC
System BIOS cacheable = disabled
Video BIOS cacheable = disabled
8 Bit I/O Recovery time = 1
16 Bit I/O Recovery time = 1
Memory Hole at 15M-16M = Disabled
Passive Release = Enabled
Delayed Transaction = Enabled
AGP Aperture Size (MB) = 64 (already fiddled with this)
CPU Clock Frequency = Default
Virus Warning = disabled
CPU L1 cache = enabled
CPU L2 cache = enabled
CPU L2 cache ECC Checking = Enabled
Gate A20 Option = Fast
PCI/VGA Palette Snoop = Enabled
OS Select for DRAM > 64 MB = Non-OS2
HDD SMART Capability = Enabled
All Video BIOS and memory address shadows are disabled
IDE HDD block mode = enabled
All PIOs and UMDAs are on auto.
On-chip primary PCI IDE amd secondary PCI IDEs are enabled
USB keyboard support = diabled (don't have a USB anything!)
Init Display First = AGP (Matrox G400 duh)

Onboard Serial Port 2 = 2F8/IRQ3
UART2 Mode Select = normal
Onboard parallel port = 378/IRQ7
Parallel Port Mode = SPP

I got an informative reply from a guy on msnews server on how to determine if L2 cache is the problem or memory. I will try those as well! So many things to do, so little time!

[antdude]

Check this post out. I got new ideas ... I think #4 is the big issue for my dilemma. I will obviously do #5 for testing. Wee so many things!

Newsgroups: microsoft.public.win98.gen_discussion
References:

Hi, Ant. There are several possibilities for your problem:

1. Re-enable the L2 cache and *enable* the L2 cache ECC.

The above will allow you to check whether the L2 cache
inside the CPU is defective or not. If you get blue-screen
errors telling you the ECC has detected memory errors
in the L2 cache memory -- then the L2 cache is at least
intermittent and the CPU should be replaced.

Note: This is a very-low-probability item. L2 cache in
Coppermine (Pentium III) chips is *very* thoroughly
tested by Intel before the chip leaves the fab.

2. If enabling the L2 cache ECC solves the problem and
the memory errors do not return, then the ECC is changing
the cache-timing in such a manner as to allow the two
memory modules to work together properly. This is not
unusual, but you may wish to update the flashBIOS to
see if you can get reliable operation with ECC disabled.
(See Item 6 for more details.)

3. If enabling the L2 cache ECC changes nothing, then the
ECC is irrelevant to your problem and you need to look
elsewhere.

4. It is *very* common for memory-module-mismatch to cause
problems with writeback from the L2 cache memory to
main memory. The usual cause for this problem is different
timing in the two memory modules that is too subtle for
the BIOS to detect and compensate for automatically.

Consequently, the next thing to do is confirm that both
memory modules work properly independently.

You know your old module is reliable -- because it worked
properly for months before you got your new one.

So, take out the old module and install the new module in
the slot where the old module was located. Do *not* just
take out the old module. Move the new module where
the old module was located.

Run for a few days with just the new module in place. If
things screw up with only the new module in place -- it is
defective and should be returned for replacement.

5. If both modules work properly independently, then the
next thing to do is check whether swapping the location
of the two modules will solve the problem.

With all motherboards that autodetect memory-timing,
the slowest module must *always* be installed in the first
memory slot -- as this is the slot which is probed by the BIOS
at startup to determine the global memory-timing parameters
for the entire memory pool.

All memory-access must be done no faster than the
slowest-reacting SIMM/DIMM in the pool. Since you know
that having the old module in the first slot doesn't work, try
running with the new module in the first slot and the old
module in the second slot. If this works, you've solved your
problem.

6. If you still have problems no matter whether the old module
is in the first or second slot -- but either module works properly
alone -- then you have a memory compatibility problem.

First step to fixing this is to upgrade to the latest motherboard
flashBIOS available for your machine. Many motherboard
manufacturers change BIOS parameters to improve memory
module compatibility as the manufacturer learns more about
how the BIOS and that particular motherboard interact.

7. If updating the BIOS does not fix the problem, then try a
matched set of memory where both modules are from the
same manufacturing batch.

If the above does not solve the problem, then check out
a matched set of very-high-quality memory from a
manufacturer such as Micron (Crucial Technology) or
Samsung (Corsair).

If the above does not solve the problem, then it is possible
you have an intermittently-defective chip on the motherboard
itself -- where the part of the chip which is not working
correctly is only accessed once you install the second
memory module. In that case, the only fix for your problem
is a new motherboard.

[Sisyfos]

FYI, according to Intel the ECC on PIII L2 cache cannot be disabled, i.e. either you run the L2 with ECC or you don't run L2 at all.
I can attest to this, as WCPUID or Powertweak always reports ECC as enabled no matter the bios settings (pertaining the L2 ECC that is).

Sorry if I add to your confusion

[antdude]

That explains WHY I don't see any differences with L2 ECC disabled!

[antdude]

YAY! My computer seems stable now! No reinstallation! Woohoo!

The solution was the newest/faster RAM module wanted to be to the closest slot to the CPU while the older/slower RAM module wanted to be the farthest slot! So, I have nothing in the middle memory slot.

WEIRD! I guess that #5 in the long, informative reply made sense. Thanks goodness, no reinstallation.

Let's hope the computer doesn't start turning evil on me . This has happened to me before!

[Gurm]

antdude:

Sounds like as a long-term goal you should be replacing your mobo and memory. Honestly. I'm glad you got it fixed for now (I think we all are), but it will break again. If one of your memory slots has gone flaky, and they're all picky... you will see trouble again. I'm not saying tomorrow. But maybe a new mobo should be on your christmas list, eh?

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.

[antdude]

Maybe. It seems like the memory speed differences (remember, the oldest module is two years old) affect which slots they are in. The long, technical e-mail and my friend said that.

[shrug]

I will definitely have to replace the motherboard when the next time to upgrade arrives (2 years?). So far, no problems last and overnight defragging my HDDs!

[xortam]

But wouldn't it be nice to get the performance gains of SIMD?

[antdude]

Shouldn't SIMD be autodetected by OS, applications, and games already? Ugh, I just don't feel like reinstalling! :I

Let me ask you this. What is the noticeable thing with SIMD enabled if I did reinstall? Does everything load faster? I noticed Windows 98 (original) takes about 10-15 seconds to boot up (I thought I was running ME or something). Diablo 2 loaded TOO FAST and smooth! Q3A, UT, and HL are friggin smooth. Maybe too much for my head to get motion sickness . As for applications like MS Word, Excel, and Paint Shop Pro 6, they loaded and acted faster.

I would like to hear more about this... Thanks!

[xortam]

Its up to the application developer as to how to implement SIMD exploitation: they may write a special app or .dll that utilizes the extensions or place it in common code and use run-time logic to decide if the processor supports the extensions and whether or not to execute the SIMD instructions. The installer could install only the appropriate .dll or install both .dlls (SIMD and non-SIMD exploitive) and leave the decision for run-time. Again, its up to the developer and not obvious to the casual user. Since Win98/ME incorporates IE (a SIMD exploiting app), the appropriate modules may only be installed at OS installation time. I don't know this for sure and haven't taken the time to investigate this: Perhaps you just need to update the plugins. A reinstallation removes all doubt. If you're dead set against reinstallation, at least consider reinstalling any multimedia apps and apps that might utilize SIMD extensions. I don't recall but perhaps Matrox's own TurboGL may be an example of delivering two implementations, one for SIMD/3DNow exploitation and one not (maybe this is commonized now). The effect of not installing the correct software is that you won't get the performance gains as it relates to streaming multimedia. Maybe you can hunt around the Web (including Intel) and get some more description of SIMD.

[antdude]

Hmmm... While I was trying to diagnose the crashes, I reinstalled a bunch of drivers. Does anyone know what games use SIMD extensions? I have HL, Q3A, UT, D2, and The Sims for current games.

[antdude]

Time to bring this thread back to life! Major updates, but I am stomped! BTW, I reinstalled Windows 98 (recreated vmm32.vxd file), but it didn't fix the problem!

I tried underclocking my CPU from 600 Mhz (100 Mhz) to 500 Mhz (83 Mhz), and my system was stable! So I have two ways to stablize my system:

1) Turn off L2 cache at 600 Mhz (100). Too slow!
2) Underclock to 500 Mhz (83). Slower, but not bad as the first method.

I finally got an informative newsgroup reply:

OK, I suspect your problem is misconfigured AGP. When you had
a p2-300, your FSB was 66mhz and your AGP ratio was (should'a been)
1:1 to make the AGP bus 66mhz. When you upgraded to a p3-600,
which uses a 100mhz FSB, you should'a changed the AGP ratio to
2:3, to get a 66mhz FSB.
You've been tryin to run that Matrox on a 100mhz AGP bus
That's why underclocking to 83mhz was successful, then you were only
tryin to run it at 83mhz. A lot of AGP cards will handle 83, but
very very few will do 100.
I'm not familiar with your mobo, but the AGP ratio will either
be a jumper or a bios setting. Usually, 1:1, 2:3, or auto. You
should use either auto or 2:3

--

I never heard of this one. Does this make any sense to you guys? Okay, how do I fix this to see if it is really works? The only AGP thing I see in CMOS is Aperture Size, but nothing to do with ratio. Maybe my motheboard doesn't support this. I believe the motherboard only does AGP 1x since it is from July 1998. I checked the motherboard manual and all I found was:

AGP -- the system board is equipped with 1 AGP slot. AGP is an interface design to support the high performance 3D graphic
cards. It utilizies a dedicated pipeline to access system memory for texturing, z-buffering, and alpha blending;
delivering up to 533 MB/sec. bandwidth for 3D graphic applications. AGP in this Pentium 2 processor based system board
will deliver faster and better graphics with your PC.

I am at a loss now.

[This message has been edited by antdude (edited 19 August 2000).]

[This message has been edited by antdude (edited 20 August 2000).]

[antdude]

Wait, that doesn't make any sense. That doesn't explain why if I do any intensive calculations (CPU burn in, search in registry file, etc. that doesn't require that much video usage), my Windows still crashes. I don't see any relationship between the video
card/AGP and CPU. Do you?

[bangs head on desk some more]

[HollyBerri]

Originally posted by antdude:
I don't see any relationship between the video card/AGP and CPU. Do you?

Errr, yes....

Ummm, everything that happens on your PC ultimately results in a video display. That's why no PC will boot if there isn't a (working) video card in it.

Does the newsgroup info make sense to me? Yes. I'll take a shot at explaining it (corrections welcomed).

You know about Front Side Bus, right? It's the speed your mobo runs the CPU at, but it also has an impact on the PCI bus and the AGP bus.

The P2-300 ran at a FSB of 66Mhz (with a multiplier of 4.5 to reach 300Mhz). The PCI bus runs at 33Mhz, and the AGP bus runs at 66Mhz (2x 33Mhz; that's what the "2X" means, twice the speed of the PCI bus, and why an AGP slot is considered a "fast PCI slot"-- because it is).

On a system where the FSB is 66Mhz, as it should have been on your old PII system (since I don't think you likely got such an old CPU to O/C to 100Mhz), the BIOS would have been set to tell the mobo to run the AGP bus at the same speed as the FSB: 66Mhz. This is expressed in the settings as a ratio of 1:1.

But on a system running at an FSB of 100Mhz, setting the AGP bus to run at 100Mhz is too fast; the card is not designed to give or receive data at that speed. In order to get the AGP bus back to running at the 66Mhz it wants, the BIOS must tell it to run at 2/3 of the FSB speed (2/3 of 100 is 66; you see?). This is expressed as a ratio of 2:3.

As I said, I had that mobo, if I had it now, I'd look it up for you, but I don't... but as far as I recall, the setting you're looking for is in the "Chipset Features" menu of the BIOS.

Good luck, hope this helps.

-------------------------------
Holly

[This message has been edited by HollyBerri (edited 20 August 2000).]

[antdude]

Interesting. I didn't think AGP, 100 Mhz and speed would relate to each other even if it didn't have to display anything on my screen (not even a single pixel). This is probably the source of all my problems [yeah right ... said that zillion of times already hehe].

As for the chipset features and settings in CMOS, I listed what I saw (I printed the screens out):

SDRAM RAS-to-CAS delay = 3
SDRAM RAS PreCharge time = 3
SDRAM CAS Latency Time = 3
SDRAM PreCharge Control = Enabled
DRAM Data Integrity Mode = Non-ECC
System BIOS cacheable = disabled
Video BIOS cacheable = disabled
Video RAM cacheable = disabled
8 Bit I/O Recovery time = 1
16 Bit I/O Recovery time = 1
Memory Hole at 15M-16M = Disabled
Passive Release = Enabled
Delayed Transaction = Enabled
AGP Aperture Size (MB) = 64
CPU Clock Frequency = Default

On the right column is the system health monitor that will beep me so I won't list column. See anything? I don't! Only the aperture size part.

It looks like I will need a new motherboard if I can't fix this, huh? Or maybe get a P3 500. Would that fix my problem if I don't want to replace my motherboard?

[This message has been edited by antdude (edited 20 August 2000).]

[This message has been edited by antdude (edited 20 August 2000).]